padosoft / askmydocs-connector-zendesk

v1.0.0 2026-10-11 19:20 UTC

Requires

• php: ^8.3
• ext-curl: *
• ext-fileinfo: *
• ext-zip: *
• illuminate/contracts: ^12.0|^13.0
• illuminate/database: ^12.0|^13.0
• illuminate/http: ^12.0|^13.0
• illuminate/queue: ^12.0|^13.0
• illuminate/support: ^12.0|^13.0
• league/html-to-markdown: ^5.1
• padosoft/askmydocs-connector-base: ^1.6

Requires (Dev)

• laravel/pint: ^1.18
• mockery/mockery: ^1.6
• orchestra/testbench: ^10.0|^11.0
• phpstan/phpstan: ^2.0
• phpunit/phpunit: ^11.0|^12.0

Apache-2.0

• Padosoft

• laravel
• Zendesk
• connectors
• knowledge-base
• customer-service
• rag
• askmydocs

Published on Packagist.org on 2026-10-11 19:51 UTC

README

Zendesk connector for AskMyDocs — tickets, comments, attachments, customers, organizations and Help Center articles, plus read-only live tools for the chat.
Drop-in Laravel package. Install it next to askmydocs-connector-base and the Zendesk card appears in the admin UI.

Table of contents

1. Why this package
2. Features
3. Architecture at a glance
4. Installation
5. Credential setup (step by step)
6. Activation inside AskMyDocs
7. What gets ingested
8. Sync semantics
9. Live tools for the chat
10. Agent usage profile
11. Trust, privacy and access control
12. Configuration reference
13. Queues and operations
14. Testing and development
15. Troubleshooting
16. Documentation index
17. License

Why this package

A customer-service desk is one of the richest sources of company knowledge: every resolved ticket is a worked example, every Help Center article is an approved answer, and every customer and organization profile is context an agent would otherwise have to look up by hand. AskMyDocs is an enterprise RAG and canonical-knowledge platform; this package pipes a Zendesk Support account into it.

It does two complementary things:

• Indexed knowledge — a resumable, checkpointed import turns tickets (with their comments), attachments, customers, organizations and published Help Center articles into KB documents, so the chat can find "how did we resolve this before?" semantically.
• Live answers — eleven read-only tools let the chat ask Zendesk right now about a specific ticket, customer, organization or article, so "what is the status of ticket #4821?" is never answered from a stale snapshot.

It follows the same contract as the sibling connectors (askmydocs-connector-freshdesk, -jira, -confluence, -google-drive, …): it extends Padosoft\AskMyDocsConnectorBase\BaseConnector, is discovered through composer.json extra.askmydocs.connectors, and never touches host classes — everything host-side goes through ConnectorIngestionContract.

Features

• 🔌 Zero-config discovery — Laravel auto-discovery registers the service provider; the connector is registered under the key zendesk and also added to connectors.built_in, so it works with a Composer path repository during development.
• 🔐 Two authentication modes — API token ({admin email}/token + token, HTTP Basic) or an OAuth access token (Bearer). Secrets live only in the encrypted credential vault, never in config_json.
• 🎫 Tickets with their conversation — subject, status, priority, type, requester, assignee, organization, channel, tags and every comment, with author names resolved.
• 🔒 Internal notes are a switch — private agent comments are included by default and can be excluded; changing the setting re-imports already imported tickets.
• 📎 Attachments as documents — PDF, DOCX, TXT and Markdown (images too when the host enables OCR). Size, count, malware-scan result, real MIME and DOCX structure are checked; credentials are sent only to the account host.
• 👤 Customers and 🏢 organizations — one profile document each (end users only; agents are never imported as customers).
• 📚 Help Center articles — published articles only; drafts are never imported and a draft removes the existing document. Optional locale filter.
• ♻️ Resumable, idempotent scans — durable checkpoints, per-installation lock, content fingerprints, Retry-After honoured, deletions only on confirmed evidence.
• 🧭 Cursor-based incremental exports — tickets and users use Zendesk's cursor-based incremental exports, the documented way to walk large accounts without skipping records.
• 🛠️ Eleven read-only live tools — opt-in per installation, project-scoped, size-capped, with explicit coverage and pagination metadata.
• 🧠 Agent usage profile — resources/agent-usage.json tells the orchestrator when to go live, when to use the index and when to combine both.
• 🏷️ Provenance and ACL reporting — implements DeclaresProvenance (tickets are customer-authored, so the default is untrusted-external) and SupportsSourceAcl (agents-only records, public or segment-restricted articles).
• 🏢 Multi-account, multi-tenant — several Zendesk accounts per tenant, each bound to its own KB project; every query is tenant-scoped.

Architecture at a glance

                ┌──────────────────────────────┐
Composer        │ padosoft/askmydocs-connector-│
require ───────▶│ zendesk (this package)       │
                └──────────────┬───────────────┘
                               │ extra.askmydocs.connectors
                               ▼
                ┌──────────────────────────────┐
                │ askmydocs-connector-base ^1.6│
                │ ConnectorRegistry · Vault ·  │
                │ SyncScheduler · TenantContext│
                └──────────────┬───────────────┘
                               │ resolves ZendeskConnector
                               ▼
   syncFull / syncIncremental ─▶ SyncManager::start()   (queues a scan, returns at once)
                                      │
                                      ▼  ProcessSyncBatch (queue "zendesk", one bounded batch per job)
        ┌──────────── phase machine, state in zendesk_sync_runs.checkpoint ────────────┐
        │ tickets → users → organizations → articles → article_check → customer_check  │
        └───────────────┬──────────────────────────────────────────────┬───────────────┘
                        │ ZendeskClient (read-only, capped, retried)    │ DocumentImporter
                        ▼                                               ▼
              https://{subdomain}.zendesk.com/api/v2        ConnectorIngestionContract (host)
                                                            ├─ redactContent()   PII at the boundary
                                                            ├─ dispatchIngestion()
                                                            └─ softDeleteByRemoteId()

   chat ──▶ ZendeskTools::catalog() / execute()  ──▶ ZendeskClient (live, read-only)

Key design decisions (full rationale in docs/architecture.md):

1. Scheduling is separate from importing. syncFull() and syncIncremental() only queue or resume a scan; the work happens in ProcessSyncBatch jobs. The $since argument of syncIncremental() is deliberately ignored — the package's own durable checkpoints are the watermark.
2. One write path. Every document kind goes through DocumentImporter::write(): redaction, fingerprint, stable path, source-aware metadata, reported ACL, counters.
3. The package never decides who may read a document. It reports what Zendesk says (SupportsSourceAcl); the host maps those identifiers to its own subjects.

Installation

Requirements: PHP 8.3+, Laravel 12 or 13, ext-curl, ext-fileinfo, ext-zip, and padosoft/askmydocs-connector-base ^1.6. The host must bind ConnectorIngestionContract (AskMyDocs ships the reference bridge) and run an asynchronous queue.

composer config repositories.askmydocs-connector-base vcs https://github.com/padosoft/askmydocs-connector-base
composer config repositories.askmydocs-connector-zendesk vcs https://github.com/padosoft/askmydocs-connector-zenddesk
composer require 'padosoft/askmydocs-connector-zendesk:^1.0'
php artisan migrate

The Composer package is padosoft/askmydocs-connector-zendesk; the GitHub repository slug is askmydocs-connector-zenddesk (historical spelling). The current release is v1.0.0; it has been validated with simulated HTTP only, so run a first scan against a Zendesk sandbox before pointing it at production (see docs/zendesk-api.md).

Optional publishes:

php artisan vendor:publish --tag=connector-zendesk-config   # config/connector-zendesk.php
php artisan vendor:publish --tag=connector-zendesk-assets   # public/connectors/zendesk.svg

The package ships one migration that creates two tables, loaded automatically: zendesk_sync_runs (one row per scan, with its checkpoint) and zendesk_source_states (what was last delivered for each remote object). The connector_installations / connector_credentials tables belong to connector-base.

Developing against a local checkout

# composer.local.json is git-ignored: a copy of composer.json whose `repositories`
# entry is a `path` repository pointing at your connector-base checkout.
COMPOSER=composer.local.json composer install

See docs/development.md.

Credential setup (step by step)

Full walkthrough with troubleshooting: docs/credential-setup.md. The short version:

Option A — API token (simplest)

1. In Zendesk open Admin Center → Apps and integrations → APIs → Zendesk API.
2. On the Settings tab make sure Token access is enabled.
3. Click Add API token, give it a description such as AskMyDocs, copy the token once (Zendesk will not show it again) and save.
4. Decide which agent the token acts as. An API token impersonates the agent whose email you pair it with, so use a dedicated agent (for example [e-mail]) that holds the Administrator role: the incremental export endpoints the import relies on are admin-only, and the connector refuses non-admin credentials when you save the form. Keep the account dedicated and its token secret — the connector itself only ever issues GET requests.
5. In AskMyDocs choose Connectors → Zendesk → Install and enter:
   • Zendesk subdomain — company for company.zendesk.com (the full host or https://company.zendesk.com is accepted too);
   • Authentication — API token;
   • Admin email — the dedicated administrator's email;
   • API token — the token from step 3.

Zendesk's credential format is {email_address}/token:{api_token} over HTTP Basic; the package builds it for you.

Option B — OAuth access token

Zendesk recommends OAuth over API tokens because a token can be scoped. Create an OAuth client in Admin Center → Apps and integrations → APIs → OAuth Clients, obtain an access token that carries the read scope, then choose OAuth access token in the form and paste it. The package sends it as Authorization: Bearer ….

Limitation (v1): the package consumes an access token you provide; it does not run the OAuth authorization-code dance or refresh expiring tokens. Use a non-expiring token, or re-enter the token when it expires (the installation shows an errored state and a 401 in its health check).

What the form checks

Saving the form calls GET /api/v2/users/me before anything reaches the vault. It must return a real user with the role admin: Zendesk answers 200 with a null identity to anonymous callers, and the incremental export APIs the import uses are admin-only, so anonymous, end-user and agent credentials are rejected. A failed re-configuration leaves the previous secret untouched.

Activation inside AskMyDocs

1. Open Settings → Connectors in the AskMyDocs admin UI.
2. The Zendesk card appears; click Install, fill in the credential form, choose a label (for example support-eu) and the KB project the documents should land in.
3. Open the installation's settings to review the sync window, internal notes, customers, organizations, Help Center, attachments and trust level (see Configuration reference).
4. Click Sync now, or wait for the scheduler (default cadence is set by connector-base). Fetch all starts the one-time historical import.

You can connect several Zendesk accounts (labels) to one tenant, each with its own project binding.

What gets ingested

Kind	Source	Document	Notes
ticket	Incremental ticket export + GET /tickets/{id} + GET /tickets/{id}/comments	Markdown: header facts, then every comment oldest-first with author and timestamp	Requester, assignee and organization names are resolved; internal notes are labelled and optional
attachment	content_url of comment attachments	The file itself (pdf, docx, txt, md; images when kb.ocr.enabled)	Child of its ticket; inline images are skipped
user	Incremental user export	Markdown profile: email, role, organization, phone, tags, details, notes	End users only; agents are never customers
organization	Incremental organization export	Markdown profile: domains, tags, details, notes
article	GET /help_center/articles	Markdown converted from the article HTML	Published only; drafts are never imported

Documents are written to {project}/connectors/zendesk/tenant-{sha256(tenant)}/installation-{id}/{kind}/{remote id}.{ext} and carry, under metadata.converter_hints.zendesk, the vendor fields (status, priority, ids, tags, timestamps, …) plus the standard _derived block (search_tags, status_active, recency_bucket). The remote identifier is stored as metadata.zendesk_remote_id = "{installation}:{kind}:{id}" — the key used for deletions and for readAcl().

Text bodies and titles pass through ConnectorIngestionContract::redactContent() before they are written or dispatched, so the host's PII policy applies to tickets, comments, profiles and articles. Binary attachments are not redacted (regex redaction cannot inspect PDFs or images); see docs/security.md.

Sync semantics

Two scan modes share one pipeline:

Mode	Start	Ticket / user / organization period
Ordinary (window)	SyncManager::start($installation) — also what syncFull() and syncIncremental() call	First scan: the last date_window_days (default 90). Later scans continue from the previous completed scan, minus a 120-second overlap, as long as the window did not grow
Historical (history, Fetch all)	SyncManager::start($installation, history: true)	From 1970-01-01 up to the scan start, regardless of the ordinary window

Help Center articles are enumerated in full in both modes (they are not windowed).

Phases run in a fixed order and skip what you disabled: tickets → users → organizations → articles → article_check, plus customer_check in history mode. State lives in zendesk_sync_runs.checkpoint; a failed run resumes where it stopped.

• Cursor-based exports. Tickets and users are read from Zendesk's cursor-based incremental exports (start_time, then the returned after_cursor until end_of_stream). A cursor that does not advance fails the scan loudly instead of looping.
• Idempotency. Each document has a SHA-256 fingerprint of its body, fields and reported ACL. Unchanged content is not delivered again; a changed ACL is.
• Rate limits. A 429 keeps the checkpoint and re-queues the batch after Retry-After seconds (60 if absent). Incremental exports are limited by Zendesk to 10 requests per minute (30 with the High Volume add-on) — the batch loop is small by design.
• Deletions need evidence.
  • A ticket is removed (with its attachments) when the export reports it with status deleted. A 404 on a ticket is not evidence (archived or inaccessible tickets answer 404).
  • A customer is removed when the export reports them as non-end-user or inactive; in history mode, customers and organizations the full export did not return are verified with show_many and removed if Zendesk omits them.
  • An article is removed when it becomes a draft, or when the article detail answers 404 after it vanished from the list. Absence from a list alone never deletes.
  • An attachment Zendesk later marks as deleted or malware is removed; one that is merely over a size or count limit is not.
  • Tickets that fall out of the ordinary window are kept.
• Settings that change content or destination (internal notes, attachment options, OCR, the KB project binding) are hashed; on change the next scan first re-imports every known ticket, customer and organization.
• Disabling a content type (customers, organizations, Help Center) stops importing it and its history reconciliation, but does not delete what was already imported.
• Rebinding to another project re-delivers every known object to the new project through that refresh pass and retires the copy in the previous one.

Details and failure semantics: docs/sync-semantics.md.

Live tools for the chat

Tools\ZendeskTools exposes a catalog and an executor, independent of the importer. For every active installation bound to the requested project that an operator has explicitly enabled (installation setting Expose live Zendesk tools to the chat, live_tools.enabled, off by default) there are eleven read-only tools, named zendesk_{installationId}_{action} (for example zendesk_42_get_ticket):

Action	Arguments	Purpose
list_tickets	one of requester_id, organization_id, email — plus cursor?, page?, per_page?	Tickets of a verified requester or organization, newest first, all statuses; without a selector, the most recently updated tickets
search_tickets	query, page?	Zendesk search syntax (free text + filters such as status:open, priority>normal, assignee_id:123); type:ticket is forced; max 1000 results
search_users	term, page?	Resolve a customer (end user) by name or email; agents are not returned
get_user	id	A user (customer or agent) by ID
search_agents	term, page?	Resolve a support agent (role:agent)
search_organizations	term	Resolve an organization by name (autocomplete)
get_organization	id	An organization by ID
get_ticket	id	A ticket by ID
list_comments	id, cursor?, per_page?	A page of comments; internal notes follow the installation setting
search_articles	term, page?	Published Help Center articles by keyword
get_article	id	A published article by ID

use Padosoft\AskMyDocsConnectorZendesk\Tools\ZendeskTools;

// The host middleware or job has already activated the right tenant.
$tools   = app(ZendeskTools::class);
$catalog = $tools->catalog('support');                       // project key is mandatory
$result  = $tools->execute('zendesk_42_get_ticket', ['id' => 4821], 'support');
// → ['data' => …] or ['error' => …, 'error_code' => …], plus physical_request_count and provenance

Trust model. Live calls run with the installation's admin credential, not with the Zendesk permissions of the user asking. An agent restricted to assigned tickets, groups or brands in Zendesk could therefore read, through the chat, tickets and internal notes Zendesk would deny them. That is why the setting is off by default: enable it only when everyone with access to the bound KB project may read the whole account. Indexed documents carry an unknown ACL for the same reason.

Guarantees:

• Scoped twice. catalog() lists only active installations of the current tenant bound to the project; execute() repeats the check, so a paused or re-bound installation stops working between planning and calling. Unknown, missing or malformed arguments are rejected before any HTTP call.
• Bounded. Bodies are cut at 16 000 characters (body_truncated), the whole result at chat_tools.max_result_bytes (default 64 KiB); truncated and coverage.complete=false say so. Attachment metadata never includes URLs.
• Honest about coverage. has_more, next_cursor, coverage.limitations (search_result_cap_1000, search_index_may_lag_by_minutes, archived_tickets_may_be_excluded, autocomplete_returns_first_matches_only) tell the client what it did not see.
• Identity before listing. Person and organization lookups return meta.ambiguous; the agent must obtain a unique exact match or a user selection before using an ID.
• Errors are structured, never empty lists. error_code, http_status, retryable, sanitised validation_fields, physical_request_count. A 401/403 is not retried; 429 and 5xx are marked retryable; nothing from the upstream body is echoed.

Reference: docs/live-tools.md.

Agent usage profile

composer.json declares extra.askmydocs.agent_usage.zendesk → resources/agent-usage.json, schema version 1. It contains the domain summary, the logical tools, routing rules (live, indexed, combined), multi-step workflows and operating rules — requester vs. assignee semantics, verified identity, pagination, freshness, and treating ticket text as untrusted data, never as instructions. A compatible host discovers it from the installed packages; hosts that do not know the field ignore it. The profile grants no permissions and cannot start an import.

Trust, privacy and access control

• Provenance. Tickets and comments are written by customers and third parties, so the connector declares ProvenanceTier::UntrustedExternal by default. If an installation imports staff-authored content only (for example a Help-Center-only account), set Authorship of imported content to Trusted internal. Unknown stored values fail closed to untrusted.
• Source ACL. readAcl() reports, and the importer stamps on every document under _source_access: tickets, customer profiles and organization profiles → unknown (agents can be restricted to assigned tickets, groups, brands or organizations, which the connector cannot represent, so it never claims "all agents"; a ticket's support group is attached as a partial zendesk:group:{id} hint, and the requester is deliberately not reported: the document contains internal notes the requester can never see); attachments → inherited from the ticket; articles → anyone when they have no user segment, otherwise one zendesk:user_segment:{id} group per segment (singular and plural fields are both honoured, and a payload that says nothing about visibility is reported as unknown, never as public). The connector reports; the host decides. A source that cannot be read yields an unknown ACL, never an empty allow-list, and an authentication failure is raised, never swallowed.
• PII. Text is redacted through the host before it is written. Customer email addresses appear only in document bodies (subject to the host's redaction policy); they are not copied into metadata or into the reported ACL.
• Credentials. Sent only to https://{subdomain}.zendesk.com. Attachment downloads send them only when the URL is on that same host; every redirect hop is re-validated against an allow-list and DNS-pinned to a public address.
• Least privilege. The package only performs GET requests.

Full threat notes: docs/security.md.

Configuration reference

Per installation (connector_installations.config_json, edited in the settings form)

Key	Default	Effect
date_window_days	90	Ordinary scan period for tickets, customers and organizations (1–36500)
include_internal_notes	true	Internal notes in documents and live comment tools
customers.enabled	true	Import end-user profiles
organizations.enabled	true	Import organization profiles
help_center.enabled	true	Import published articles
help_center.locale	(empty)	Restrict articles to one locale such as en-us; empty = every locale the agent can read
attachments.enabled	true	Download compatible attachments
attachments.max_size_mb	25	Per-file limit (1–100 MiB)
attachments.max_per_ticket	20	Compatible files per ticket (1–100)
provenance_tier	untrusted-external	trusted-internal or untrusted-external
connection.subdomain / auth_mode / email	—	Set by the credential form; the secret is in the vault

Package (config/connector-zendesk.php)

Key	Default	Effect
http.timeout / connect_timeout	30 / 5 s	API timeouts
http.attempts	3	Tries for connection failures and 5xx
http.max_json_bytes	16 000 000	Cap on one API response
http.resolve_dns	true	DNS-pin public addresses; false is for simulated HTTP in tests only
attachments.allowed_hosts	*.zendesk.com, *.zdusercontent.com, *.zdassets.com, S3 hosts	Where attachments (and their redirects) may come from
sync.batch_items / batch_seconds	5 / 60	Steps and indicative duration of one batch
sync.overlap_seconds	120	Re-scan overlap before the previous watermark
sync.export_min_age_seconds	90	Keeps the export start_time safely in the past
sync.ticket_support_type_scope	(unset)	CONNECTOR_ZENDESK_TICKET_SCOPE: agent, ai_agent or all, sent as support_type_scope to the ticket export. Unset keeps Zendesk's default; use all to include AI-agent tickets (not yet verified live). Changing it restarts the ticket window on the next scan
attachments.allow_unscanned	false	CONNECTOR_ZENDESK_ALLOW_UNSCANNED: also import files whose malware scan is not_scanned, failed_to_scan or missing. Off by default: only malware_not_found is imported
sync.queue / sync.connection	zendesk / (default)	CONNECTOR_ZENDESK_QUEUE, CONNECTOR_ZENDESK_CONNECTION
sync.middleware	[]	Job middleware for ProcessSyncBatch and StartSync
chat_tools.enabled	true	Global kill switch: false removes every live tool. Tools also need the per-installation live_tools.enabled setting (off by default)
chat_tools.max_result_bytes	65536	Cap on one tool result

Queues and operations

php artisan queue:work zendesk --queue=zendesk --timeout=600 --tries=3

• Use an asynchronous queue (a synchronous queue.default or CONNECTOR_ZENDESK_CONNECTION is refused) and a cache store with atomic locks (Redis): one batch per installation runs at a time, others are deferred by 30 seconds.
• The queue connection's retry_after must be greater than 600 seconds.
• StartSync($installationId, $tenantId) is the entry point for your own scheduler; it queues or resumes a scan without marking anything complete.
• A worker timeout leaves a resumable failure (the installation shows errored) and keeps the checkpoint; starting a scan again resumes it. No exception body is ever persisted.
• Pausing or disconnecting an installation cancels its scan at the next step.

Testing and development

composer install            # or: COMPOSER=composer.local.json composer install
composer test               # PHPUnit + Testbench, SQLite, simulated HTTP
composer analyse            # PHPStan level 6
vendor/bin/pint --test      # code style

The suite covers credential verification and state tokens, both auth modes, full scans and idempotency, internal-note and content-type switches, deletions on evidence only, attachments (credentials scoping, redirects, MIME, size, malware flag), rate limits, resume after failure, locks, cancellation, tenant isolation, every live tool, the agent profile, provenance and ACL. CI runs PHP 8.3/8.4 × Laravel 12/13. No test talks to a real Zendesk account; see Troubleshooting and docs/zendesk-api.md for what has and has not been verified against a live instance.

Troubleshooting

Symptom	Likely cause	Fix
Install says Zendesk did not recognise the credentials	users/me returned an anonymous identity: wrong token, wrong subdomain, or token access disabled	Re-check the token, the admin email and Token access in Admin Center
Use an admin account	The email belongs to an agent or end user	Pair the token with an admin email
Zendesk rejected the credentials (401) in health	Token deleted/rotated, or OAuth token expired	Create a new token and re-enter credentials
The Zendesk account does not permit this operation (403)	The agent's role cannot read that resource (for example users or Help Center)	Grant the role access, or switch the content type off
Scan stays queued	No worker on the zendesk queue, or a sync queue	Run the worker command above; use an async queue
Zendesk rate limit reached	Incremental exports are limited per minute	Nothing to do: the batch resumes after Retry-After
export cursor did not advance	Zendesk returned the same cursor without ending the stream	Re-run; if persistent, open an issue with the run's checkpoint
Ticket appears in search but not in the index	It is outside date_window_days	Widen the window or use Fetch all
New ticket missing from live search	Zendesk search is indexed with a delay of minutes	Use get_ticket, or retry shortly
Internal notes still visible after switching off	The re-import runs on the next scan	Trigger Sync now
Deleted customer still in the KB	Ordinary scans only see exports inside the window	Run Fetch all once; it verifies absent customers

More: docs/troubleshooting.md.

Documentation index

Page	Contents
docs/architecture.md	Components, data flow, design decisions and trade-offs
docs/credential-setup.md	Creating tokens/OAuth credentials, roles, rotation
docs/sync-semantics.md	Phases, checkpoints, deletions, failure and resume behaviour
docs/live-tools.md	Tool reference, argument rules, result and error shapes
docs/data-model.md	Tables, document layout, metadata, remote identifiers
docs/security.md	Threat model, SSRF/credential handling, PII, ACL, provenance
docs/zendesk-api.md	Every endpoint used, with its verification status
docs/troubleshooting.md	Symptom → cause → fix
docs/development.md	Local setup, tests, extending the package
CHANGELOG.md	Release notes

License

Apache-2.0 — see LICENSE.

Zendesk is a trademark of Zendesk, Inc. This project is independent and is not endorsed by or affiliated with Zendesk. See public/icons/README.md for the icon's provenance.

Built and maintained by Padosoft. Part of the AskMyDocs connector ecosystem.